AAU Student Projects - visit Aalborg University's student projects portal
A master's thesis from Aalborg University
Book cover


Time Predictable Dalvik v0.1: An analysis of the DVM opcodes with focus on timing predictability

Author

Term

4. term

Education

Publication year

2015

Pages

30

Abstract

Dette projekt analyserer, hvordan hver enkelt opcode (de lavniveau-instruktioner, som en virtuel maskine udfører) er implementeret i Dalvik Virtual Machine (DVM). Analysen fokuserer på den portable udgave af den iterative fortolker (som udfører instruktioner én ad gangen), med særligt fokus på tidsmæssig forudsigelighed (hvor forudsigeligt køretiden kan bestemmes). For hver gennemgået implementering skitserer vi, hvordan den kan omstruktureres (refaktoreres). Afslutningsvis diskuterer vi, hvad der yderligere kræves for at gøre DVM fuldt tidsforudsigelig, herunder behovet for en anden programmeringsmodel samt at erstatte eller fjerne den underliggende Linux-kerne.

This project analyzes how each opcode—the low-level instructions a virtual machine executes—is implemented in the Dalvik Virtual Machine (DVM). The analysis focuses on the portable version of the iterative interpreter (which runs instructions one by one), with special attention to timing predictability (how reliably execution time can be anticipated). For each implementation we review, we outline a way to refactor it. Finally, we discuss what additional changes would be needed to make the DVM fully timing-predictable, including the need for a different programming model and the replacement or removal of the underlying Linux kernel.

[This abstract was generated with the help of AI]